使用 `Roslyn` 分析器和修复器对.cs源代码添加头部注释

之前写过两篇关于Roslyn源生成器生成源代码的用例,今天使用Roslyn的代码修复器CodeFixProvider实现一个cs文件头部注释的功能,

代码修复器会同时涉及到CodeFixProvider和DiagnosticAnalyzer,

实现FileHeaderAnalyzer

首先我们知道修复器的先决条件是分析器,比如这里,如果要对代码添加头部注释,那么分析器必须要给出对应的分析提醒:

我们首先实现实现名为FileHeaderAnalyzer的分析器:

[DiagnosticAnalyzer(LanguageNames.CSharp)]
public class FileHeaderAnalyzer : DiagnosticAnalyzer
{
    public const string DiagnosticId = "GEN050";
    private static readonly LocalizableString Title = "文件缺少头部信息";
    private static readonly LocalizableString MessageFormat = "文件缺少头部信息";
    private static readonly LocalizableString Description = "每个文件应包含头部信息.";
    private const string Category = "Document";

    private static readonly DiagnosticDescriptor Rule = new(
        DiagnosticId, Title, MessageFormat, Category, DiagnosticSeverity.Warning, isEnabledByDefault: true, description: Description);

    public override ImmutableArray<DiagnosticDescriptor> SupportedDiagnostics => [Rule];

    public override void Initialize(AnalysisContext context)
    {
        if (context is null)
            return;

        context.ConfigureGeneratedCodeAnalysis(GeneratedCodeAnalysisFlags.None);
        context.EnableConcurrentExecution();
        context.RegisterSyntaxTreeAction(AnalyzeSyntaxTree);
    }

    private static void AnalyzeSyntaxTree(SyntaxTreeAnalysisContext context)
    {
        var root = context.Tree.GetRoot(context.CancellationToken);
        var firstToken = root.GetFirstToken();

        // 检查文件是否以注释开头
        var hasHeaderComment = firstToken.LeadingTrivia.Any(trivia => trivia.IsKind(SyntaxKind.SingleLineCommentTrivia) || trivia.IsKind(SyntaxKind.MultiLineCommentTrivia));

        if (!hasHeaderComment)
        {
            var diagnostic = Diagnostic.Create(Rule, Location.Create(context.Tree, TextSpan.FromBounds(0, 0)));
            context.ReportDiagnostic(diagnostic);
        }
    }
}

FileHeaderAnalyzer分析器的原理很简单,需要重载几个方法,重点是Initialize方法,这里的RegisterSyntaxTreeAction即核心代码,SyntaxTreeAnalysisContext对象取到当前源代码的SyntaxNode根节点,然后判断TA的第一个SyntaxToken是否为注释行(SyntaxKind.SingleLineCommentTrivia|SyntaxKind.MultiLineCommentTrivia)

如果不为注释行,那么就通知分析器!

实现了上面的代码我们看一下效果:

并且编译的时候分析器将会在错误面板中显示警告清单:

实现CodeFixProvider

分析器完成了,现在我们就来实现名为AddFileHeaderCodeFixProvider的修复器,

/// <summary>
/// 自动给文件添加头部注释
/// </summary>
[ExportCodeFixProvider(LanguageNames.CSharp, Name = nameof(AddFileHeaderCodeFixProvider))]
[Shared]
public class AddFileHeaderCodeFixProvider : CodeFixProvider
{
    private const string Title = "添加文件头部信息";
    //约定模板文件的名称
    private const string ConfigFileName = "Biwen.AutoClassGen.Comment";
    private const string VarPrefix = "$";//变量前缀
    //如果模板不存在的时候的默认注释文本
    private const string DefaultComment = """
        // Licensed to the {Product} under one or more agreements.
        // The {Product} licenses this file to you under the MIT license.
        // See the LICENSE file in the project root for more information.
        """;

    #region regex

    private const RegexOptions ROptions = RegexOptions.Compiled | RegexOptions.Singleline;
    private static readonly Regex VersionRegex = new(@"<Version>(.*?)</Version>", ROptions);
    private static readonly Regex CopyrightRegex = new(@"<Copyright>(.*?)</Copyright>", ROptions);
    private static readonly Regex CompanyRegex = new(@"<Company>(.*?)</Company>", ROptions);
    private static readonly Regex DescriptionRegex = new(@"<Description>(.*?)</Description>", ROptions);
    private static readonly Regex AuthorsRegex = new(@"<Authors>(.*?)</Authors>", ROptions);
    private static readonly Regex ProductRegex = new(@"<Product>(.*?)</Product>", ROptions);
    private static readonly Regex TargetFrameworkRegex = new(@"<TargetFramework>(.*?)</TargetFramework>", ROptions);
    private static readonly Regex TargetFrameworksRegex = new(@"<TargetFrameworks>(.*?)</TargetFrameworks>", ROptions);
    private static readonly Regex ImportRegex = new(@"<Import Project=""(.*?)""", ROptions);

    #endregion

    public sealed override ImmutableArray<string> FixableDiagnosticIds
    {
        //重写FixableDiagnosticIds,返回分析器的报告Id,表示当前修复器能修复的对应Id
        get { return [FileHeaderAnalyzer.DiagnosticId]; }
    }

    public sealed override FixAllProvider GetFixAllProvider()
    {
        return WellKnownFixAllProviders.BatchFixer;
    }

    public override Task RegisterCodeFixesAsync(CodeFixContext context)
    {
        var diagnostic = context.Diagnostics[0];
        var diagnosticSpan = diagnostic.Location.SourceSpan;

        context.RegisterCodeFix(
            CodeAction.Create(
                title: Title,
                createChangedDocument: c => FixDocumentAsync(context.Document, diagnosticSpan, c),
                equivalenceKey: Title),
            diagnostic);

        return Task.CompletedTask;
    }

    private static async Task<Document> FixDocumentAsync(Document document, TextSpan span, CancellationToken ct)
    {
        var root = await document.GetSyntaxRootAsync(ct).ConfigureAwait(false);

        //从项目配置中获取文件头部信息
        var projFilePath = document.Project.FilePath ?? "C:\\test.csproj";//单元测试时没有文件路径,因此使用默认路径

        var projectDirectory = Path.GetDirectoryName(projFilePath);
        var configFilePath = Path.Combine(projectDirectory, ConfigFileName);

        var comment = DefaultComment;

        string? copyright = "MIT";
        string? author = Environment.UserName;
        string? company = string.Empty;
        string? description = string.Empty;
        string? title = document.Project.Name;
        string? version = document.Project.Version.ToString();
        string? product = document.Project.AssemblyName;
        string? file = Path.GetFileName(document.FilePath);
        string? targetFramework = string.Empty;
#pragma warning disable CA1305 // 指定 IFormatProvider
        string? date = DateTime.Now.ToString("yyyy-MM-dd HH:mm:ss");
#pragma warning restore CA1305 // 指定 IFormatProvider

        if (File.Exists(configFilePath))
        {
            comment = File.ReadAllText(configFilePath, System.Text.Encoding.UTF8);
        }

        #region 查找程序集元数据

        // 加载项目文件:
        var text = File.ReadAllText(projFilePath, System.Text.Encoding.UTF8);
        // 载入Import的文件,例如 : <Import Project="..\Version.props" />
        // 使用正则表达式匹配Project:
        var importMatchs = ImportRegex.Matches(text);
        foreach (Match importMatch in importMatchs)
        {
            var importFile = Path.Combine(projectDirectory, importMatch.Groups[1].Value);
            if (File.Exists(importFile))
            {
                text += File.ReadAllText(importFile);
            }
        }

        //存在变量引用的情况,需要解析
        string RawVal(string old, string @default)
        {
            if (old == null)
                return @default;

            //当取得的版本号为变量引用:$(Version)的时候,需要再次解析
            if (version.StartsWith(VarPrefix, StringComparison.Ordinal))
            {
                var varName = old.Substring(2, old.Length - 3);
                var varMatch = new Regex($@"<{varName}>(.*?)</{varName}>", RegexOptions.Singleline).Match(text);
                if (varMatch.Success)
                {
                    return varMatch.Groups[1].Value;
                }
                //未找到变量引用,返回默
                return @default;
            }
            return old;
        }

        var versionMatch = VersionRegex.Match(text);
        var copyrightMath = CopyrightRegex.Match(text);
        var companyMatch = CompanyRegex.Match(text);
        var descriptionMatch = DescriptionRegex.Match(text);
        var authorsMatch = AuthorsRegex.Match(text);
        var productMatch = ProductRegex.Match(text);
        var targetFrameworkMatch = TargetFrameworkRegex.Match(text);
        var targetFrameworksMatch = TargetFrameworksRegex.Match(text);

        if (versionMatch.Success)
        {
            version = RawVal(versionMatch.Groups[1].Value, version);
        }
        if (copyrightMath.Success)
        {
            copyright = RawVal(copyrightMath.Groups[1].Value, copyright);
        }
        if (companyMatch.Success)
        {
            company = RawVal(companyMatch.Groups[1].Value, company);
        }
        if (descriptionMatch.Success)
        {
            description = RawVal(descriptionMatch.Groups[1].Value, description);
        }
        if (authorsMatch.Success)
        {
            author = RawVal(authorsMatch.Groups[1].Value, author);
        }
        if (productMatch.Success)
        {
            product = RawVal(productMatch.Groups[1].Value, product);
        }
        if (targetFrameworkMatch.Success)
        {
            targetFramework = RawVal(targetFrameworkMatch.Groups[1].Value, targetFramework);
        }
        if (targetFrameworksMatch.Success)
        {
            targetFramework = RawVal(targetFrameworksMatch.Groups[1].Value, targetFramework);
        }

        #endregion

        //使用正则表达式替换
        comment = Regex.Replace(comment, @"\{(?<key>[^}]+)\}", m =>
        {
            var key = m.Groups["key"].Value;
            return key switch
            {
                "Product" => product,
                "Title" => title,
                "Version" => version,
                "Date" => date,
                "Author" => author,
                "Company" => company,
                "Copyright" => copyright,
                "File" => file,
                "Description" => description,
                "TargetFramework" => targetFramework,
                _ => m.Value,
            };
        }, RegexOptions.Singleline);

        var headerComment = SyntaxFactory.Comment(comment + Environment.NewLine);
        var newRoot = root?.WithLeadingTrivia(headerComment);
        if (newRoot == null)
        {
            return document;
        }
        var newDocument = document.WithSyntaxRoot(newRoot);

        return newDocument;
    }
}

代码修复器最重要的重载方法RegisterCodeFixesAsync,对象CodeFixContext包含项目和源代码以及对应分析器的信息:

比如:CodeFixContext.Document表示对应的源代码,CodeFixContext.Document.Project表示对应项目,CodeFixContext.Document.Project.FilePath就是代码中我需要的*.csproj的项目文件,

我们取到项目文件,那么我们就可以读取配置在项目文件中的信息,比如Company,Authors,Description,甚至上一篇我们提到的版本号等有用信息,当前我用的正则表达式,当然如果可以你也可以使用XPath,

然后取到的有用数据替换模板即可得到想要的注释代码片段了!

比如我的Comment模板文件Biwen.AutoClassGen.Comment

// Licensed to the {Product} under one or more agreements.
// The {Product} licenses this file to you under the MIT license.
// See the LICENSE file in the project root for more information.
// {Product} Author: 万雅虎 Github: https://github.com/vipwan
// {Description}
// Modify Date: {Date} {File}

替换后将会生成如下的代码:

// Licensed to the Biwen.QuickApi under one or more agreements.
// The Biwen.QuickApi licenses this file to you under the MIT license.
// See the LICENSE file in the project root for more information.
// Biwen.QuickApi Author: 万雅虎 Github: https://github.com/vipwan
// Biwen.QuickApi ,NET9+ MinimalApi CQRS
// Modify Date: 2024-09-07 15:22:42 Verb.cs

最后使用SyntaxFactory.Comment(comment)方法生成一个注释的SyntaxTrivia并附加到当前的根语法树上,最后返回这个新的Document即可!

大功告成,我们来看效果:

以上代码就完成了整个源生成步骤,最后你可以使用我发布的nuget包体验:

dotnet add package Biwen.AutoClassGen

源代码我发布到了GitHub,欢迎star! https://github.com/vipwan/Biwen.AutoClassGen

https://github.com/vipwan/Biwen.AutoClassGen/blob/master/Biwen.AutoClassGen.Gen/CodeFixProviders/AddFileHeaderCodeFixProvider.cs